Implementation of Reinforcement Learning Algorithms. Python, OpenAI Gym, Tensorflow. Exercises and Solutions to accompany Sutton's Book and David Silver's course.
The codes in DQN
conv1 = tf.contrib.layers.conv2d(X, 32, 8, 4, activation_fn=tf.nn.relu)
the kernel size is 8. Why even numbers are used as convolution kernels ?
and the codes:
state = env.reset()
state = state_processor.process(sess, state)
state = np.stack([state] * 4, axis=2)
Why is the number of input channels 4 ？ Why input the same 4 images instead of 1?
The codes in DQN conv1 = tf.contrib.layers.conv2d(X, 32, 8, 4, activation_fn=tf.nn.relu) the kernel size is 8. Why even numbers are used as convolution kernels ? and the codes: state = env.reset() state = state_processor.process(sess, state) state = np.stack([state] * 4, axis=2) Why is the number of input channels 4 ？ Why input the same 4 images instead of 1?
Hope you answer... @dennybritz